Repolode
Submit
Tencent

Tencent/AI-Infra-Guard

A full-stack AI red teaming platform by Tencent Zhuque Lab that scans AI agents, MCP servers, Agent Skills and AI infrastructure for security risks.

View on GitHubWebsite
Stars
6,933
Forks
670
Open issues
22
Last commit
4 hours ago

What is it?

AI-Infra-Guard (A.I.G) is a security testing platform for AI systems. It covers several scanning areas: ClawScan for OpenClaw security risks, Agent Scan for agent workflows (including Dify and Coze agents), MCP Server and Agent Skills scanning from source code or remote URLs, AI infrastructure vulnerability scanning against live services, and jailbreak evaluation of LLMs using multiple attack methods.

The AI infrastructure scanner fingerprints running services such as Ollama, ComfyUI, vLLM, n8n and Triton Inference Server, then matches them against a library of CVE rules. Targets can be a single address, a CIDR range or an IP range. The project also includes a model and API relay checker, and a skill-scan CLI that can be run in CI/CD pipelines.

A.I.G runs via Docker Compose and exposes a web interface on port 8088 with a Swagger API. The project is intended for internal use and the maintainers advise against exposing it on public networks because it lacks authentication.

Key features

  • 01Scans MCP Servers and Agent Skills from source code or remote URLs across 14 risk categories.
  • 02Identifies over 100 AI framework components and matches them against 2000+ CVE rules.
  • 03Runs multi-attack jailbreak evaluations and compares results across models.
  • 04Provides a skill-scan CLI installable with pip for CI/CD pipelines.
  • 05Offers a web UI, Swagger API docs and Docker-based deployment.

Use cases

  • Auditing an internal vLLM, Ollama or ComfyUI deployment for known CVEs before exposing it to users.
  • Checking third-party MCP servers or Agent Skills for malicious behavior before installing them.
  • Adding Agent Skill security checks to an enterprise CI/CD pipeline.
  • Running jailbreak evaluations against an LLM endpoint to compare model robustness.

Quick start

  1. 1

    Clone the repository

    git clone https://github.com/Tencent/AI-Infra-Guard.git
  2. 2

    Enter the project directory

    cd AI-Infra-Guard
  3. 3

    Start the service with Docker Compose (pulls pre-built images)

    docker-compose -f docker-compose.images.yml up -d
  4. 4

    Open the web interface at http://localhost:8088

Things to know

  • The README says it lacks authentication and should not be deployed on public networks.
  • Some features, such as the Pro version, require an invitation code.
  • Skill scanning accuracy depends on an external LLM API key.

Topics

#ai-security#red-teaming#llm-security#mcp#agent-security#vulnerability-scanner#prompt-injection#cve#agent#agentic-ai

FAQ

What is AI-Infra-Guard?
A full-stack AI red teaming platform by Tencent Zhuque Lab that scans AI agents, MCP servers, Agent Skills and AI infrastructure for security risks. AI-Infra-Guard (A.I.G) is a security testing platform for AI systems. It covers several scanning areas: ClawScan for OpenClaw security risks, Agent Scan for agent workflows (including Dify and Coze agents), MCP Server and Agent Skills scanning from source code or remote URLs, AI infrastructure vulnerability scanning against live services, and jailbreak evaluation of LLMs using multiple attack methods.
Is AI-Infra-Guard free to use? What license does it have?
AI-Infra-Guard is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
Is AI-Infra-Guard still maintained?
Actively maintained. The last commit to AI-Infra-Guard was on Oct 22, 2026.
What can I use AI-Infra-Guard for?
Auditing an internal vLLM, Ollama or ComfyUI deployment for known CVEs before exposing it to users. Checking third-party MCP servers or Agent Skills for malicious behavior before installing them. Adding Agent Skill security checks to an enterprise CI/CD pipeline. Running jailbreak evaluations against an LLM endpoint to compare model robustness.
How do I install AI-Infra-Guard?
Clone the repository: git clone https://github.com/Tencent/AI-Infra-Guard.git Enter the project directory: cd AI-Infra-Guard Start the service with Docker Compose (pulls pre-built images): docker-compose -f docker-compose.images.yml up -d Open the web interface at http://localhost:8088
Where can I get AI-Infra-Guard?
The source code and installation instructions are on GitHub: https://github.com/Tencent/AI-Infra-Guard

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects